92929 - global variable cannot be used before it is declared

Status: VERIFIED FIXED

(Core :: XSLT, defect)

Description

[Oleg Tkachenko]

Hi guys !

It's me again, excuse me, but I've found something wrong about global variables - 
they cannot be used before they are declared. Now I'll try to attach simple demo
stylesheet.

Comment 1

[Oleg Tkachenko]

Attachment: Stylesheet

Comment 2

[Oleg Tkachenko]

Attachment: Source, uses above stylesheet

Comment 3

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

yes, we are severly broken when it comes to global variables (and rather broken 
for other variables as well). We should probobly evaluate global variables 
lazily. That way we can also support import precedence correctly aswell.

We also have to watch out for infinite recursions like:

 <xsl:variable name="var1" select="$var2 + 2"/>
 <xsl:variable name="var2" select="$var1 + 2"/>

Nice testcases btw ;)

Comment 4

[Axel Hecht]

This is import precedence, too. Adding a dependency. Not sure which direction,
though ;-).
I guess we should have a step evalutateGlobalVariables between 
processTopLevel and process.

Axel

Comment 5

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

either that (there are a couple of other things we could do there too), or we 
evaluate global variables lazily.

I.e. we don't do anything when finding a variable in processTopLevel other then 
add it in the appropriate importContainer. Then when we resolve a variable, 
after we've walked up the varaible-bindings stack, we look for a global 
variable, if one exists we set it's "evaluating" flag (to prevent infinite 
recursion) and then evaluate the variable.

Comment 6

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

taking, this is fixed by my patch in bug 117658

Comment 7

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

...

Comment 8

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

fixed with the checkin from bug 117658

Comment 9

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

fixed with the branchlanding

Comment 10

[Axel Hecht]

we didn't verify for a long time.
I really checked, so VERIFIED.